INTRODUCTION

Roughly 150,000 new cases of colorectal cancer are diagnosed each year in the United States primarily through colonoscopies, considered to be the gold standard for colon cancer screening. Several quality benchmarks for colonoscopy have been proposed and widely adopted in the healthcare system. In particular, adenoma detection rate (ADR) is an important quality measure for screening colonoscopy (1). Higher ADR is associated with reduction in interval colon cancer and cancer-related mortality (2,3). Recently, overall ADR is reported to increase over time (4,5); however, the factors contributing to this increase remain unclear. Some studies showed that ADR variation is affected by patient-level characteristics such as age and sex (6–8), whereas other studies have reported physician-level factors can also influence ADR, such as sex of physician and years of training (9,10). There are limited data on the relative effect of patient, physician, and procedure factors on changes in ADR over time.

Although multiple professional societies have recommended using ADR thresholds to assess the quality of colonoscopies, there are questions regarding use of ADR as the best quality metric for prevention of postcolonoscopy colorectal cancer (1,11). Sessile serrated lesions (SSLs) are recognized precursors to 30% of colon cancer (12), but a performance threshold has yet to be established. Although sessile serrated lesions detection rate (SSLDR) has been correlated with ADR, data regarding trends of SSLDR over time are limited, making an SSLDR target difficult to gauge (1,13). In this study, we aim to (i) examine the trends in ADR and SSLDR over time and (ii) explore associations of patient-level, physician-level, and procedure-level factors with ADR and SSLDR and whether the effects of these factors change over time.

METHODS Study setting

We conducted a retrospective cohort study using data from a large integrated healthcare system in northern California, Sutter Health-Palo Alto Medical Foundation (PAMF). PAMF serves a diverse population of approximately 1 million active patients annually across 4 counties in the San Francisco Bay area. The healthcare system provides primary care, specialty care, ambulatory surgery, and laboratory, imaging, and inpatient physician services. Most patient care is recorded in an unified electronic health record system.

Study sample

Our starting sample included 192,673 total colonoscopies performed for 144,122 adult patients at the PAMF ambulatory surgical centers between January 2010 and December 2020. We excluded those younger than 18 years (n = 131), with a history of colon cancer (n = 1,212) or history of inflammatory bowel disease (n = 5,859), with diagnostic colonoscopies (n = 37,413), or with procedure performed at a non-PAMF facility (n = 8,131). This yielded a final sample of 146,818 screening colonoscopies performed by 33 endoscopists with 88,091 pathology reports (Figure 1).

Figure 1.:

Colonoscopy study sample.

Polyp outcomes

We developed a natural language processing (NLP) algorithm to extract colon polyp data (presence of tubular, tubulovillous, villous, sessile serrated, and traditional adenoma) from available free-text pathology reports. The algorithm was validated against a random sample of 1,000 manual chart reviews with more than 98% agreement rate among 2 reviewers (authors E.S.H. and S.-Y.L.) and a Cohen's kappa statistics of 0.959. We subsequently applied this NLP algorithm to the final analytic sample (see Supplementary Appendix 1, Supplementary Digital Content 1, https://links.lww.com/CTG/B84).

For each colonoscopy, we used the NLP-derived adenoma information to classify whether at least 1 adenoma (either tubular, tubulovillous, or villous) was detected. Similarly, we used the NLP-derived information to classify whether sessile serrated lesions (either sessile serrated or traditional serrated adenoma) were detected as a secondary outcome variable. The interpretation of sessile serrated lesions was captured using histologic criteria at the time of collection (14). We defined ADR as the number of colonoscopies with adenomatous polyps over total number of screening colonoscopies and SSLDR as number of colonoscopies with sessile serrated lesion over total number of screening colonoscopies.

Patient, procedure, and physician characteristics

Patient and physician data were obtained from structured electronic health record fields. Patient characteristics were collected as follows: age, sex, race, body mass index (BMI), smoking history, Charlson comorbidity index (CCI, using International Classification of Disease, 9th revision and 10th revision), and colonoscopy indication (average risk and high risk). Average risk was defined as those without any personal history of colon polyp, family history of colon cancer, and family history of colon polyps, whereas high risk was defined as having 1 or more of these characteristics. Colonoscopy elements were collected including use of mucosal assist device, use of high-definition colonoscope, cecal withdrawal time, quality of preparation, and type of sedation. Physician-level data included sex and years in practice. At the time of data extraction, approximately 23% of procedures did not have information on preparation quality and sedation type, thus missingness was imputed by random sampling of values proportionally to retain the existing distribution.

Statistical analysis

We computed the yearly ADR and SSLDR between 2010 and 2020. Linear regressions were used to examine the trends over time. χ2 tests were used to assess bivariate associations of patient factors: age (<50, 50–75, or >75 years), sex (male or female), BMI (<25, 25–30, >30 kg/m2, or unknown), race (White, Asian American, African American, other, or unknown), smoking status (never, current, former, or unknown), CCI (0, 1, or ≥2), colonoscopy type (average risk vs high risk), procedure characteristics: use of mucosal assist device (yes or no), use of high-definition colonoscope (yes, no, or missing), cecal withdrawal time (<6, 6–9, >9 minutes, or unknown), quality of preparation (adequate or inadequate), type of sedation (moderate sedation or other), and physician factors: sex (male or female) and years in practice (<11, 11–25, or >25 years), with the study outcomes. Generalized linear models were used to estimate the odds ratios (ORs) and 95% confidence interval (CI), adjusting for covariates and account for clustering of patients cared for by the same endoscopist. To explore whether risk factors have different effects on the detection of adenoma and sessile serrated lesions over time, we conducted stratified analyses focusing on the first year (2010) and the last year (2020) of the study period. We estimated the generalized linear model using the pooled 2010 and 2020 data and adding the interaction terms between the year and risk factors. Testing the effects are the same between 2010 and 2020 was performed using Wald tests to assess whether coefficients of interaction terms being zero. We discuss results as statistically significant when P < 0.05. All P values were 2-sided. Analyses were conducted using R (version 4.2.2; The R Project) and Stata (version 16.1; StataCorp LLC, College Station, TX). The study was reviewed and approved by the Sutter Health Institutional Review Board.

RESULTS Sample characteristics

Of 146,818 screening colonoscopies, the overall mean age of patients was 58.7 years (SD 9.1) and 50% were women. The mean BMI was 26.9 (SD 5.0). Approximately 55.5% were White, 26.6% Asian, 1.9% African American, 4.1% other, and 11.9% unknown. Nineteen-point-six percent were former smokers, whereas 3.5% were current smokers. The mean CCI was 1.33 (SD 1.79), and 55.8% of colonoscopies were average risk. Approximately 15.4% and 55.2% of colonoscopies used a mucosal assist device and high-definition colonoscopes, respectively. The median cecal withdrawal time was 13 minutes (interquartile range 10–17 minutes). Nearly all (98.9%) colonoscopies had adequate preparation quality, and 94.7% used moderate sedation. Seventy-five percent of procedures were performed by male physicians. The mean years of training was 15.4 (SD 7.7) (Table 1).

Table 1. - Baseline characteristics Characteristics Overall (n = 146,818) Patient-related  Age, yr, mean ± SD 58.7 ± 9.1  Sex, n (%)   Male 73,645 (50.2)   Female 73,173 (49.8)  BMI, kg/m2, mean ± SD 26.9 ± 5.0  Race, n (%)   White 81,473 (55.5)   Asian 39,059 (26.6)   African American 2,832 (1.9)   Other 5,980 (4.1)   Unknown 17,474 (11.9)  Smoking status, n (%)   Never smoker 101,419 (69.1)   Current smoker 5,126 (3.5)   Former smoker 28,785 (19.6)   Unknown 11,488 (7.8)  Charlson comorbidity, mean ± SD 1.3 ± 1.8  Colonoscopy indication, n (%)   Average risk 81,893 (55.8)   High risk 64,925 (44.2) Procedure-related  Mucosal assist device, n (%)   No 124,195 (84.6)   Yes 22,623 (15.4)  High-definition colonoscope   No 30,289 (20.6)   Yes 81,021 (55.2)   Unknown 35,508 (24.2)  Withdrawal time, min, median (IQR) 13 (10–17)  Preparation quality, n (%)   Adequate (excellent, good, or fair) 145,241 (98.9)   Inadequate (poor) 1,577 (1.1)  Sedation type, n (%)   Moderate sedation 139,056 (94.7)   Othera 7,762 (5.3) Physician-related  Procedure performed by male physician   Male physician 110,018 (74.9)   Female physician 36,800 (25.1)  Years in practice, yr, mean ± SD 15.4 ± 7.7

BMI, body mass index; IQR, interquartile range.

aOther includes no sedation or general anesthesia.

Of the 88,091 pathology reports, 63.0% had adenoma, and 13.8% had sessile serrated lesion, whereas 6.2% of the pathology contained both adenoma and sessile serrated lesion (see Supplementary Table 1, Supplementary Digital Content 1, https://links.lww.com/CTG/B84).

ADR and SSLDR: trend over time

Figure 2 describes trends of ADR and SSLDR over time. The volume of colonoscopies increased between 2010 and 2019, but dropped in 2020 during the COVID-19 pandemic. The overall ADR ranged 19.4%–22.9% in 2010–2011, 28.5%–34.0% in 2012–2016, and 45.6%–44.4% in 2017–2020. ADR increased by a mean of 2.67% each year (95% CI 1.93%–3.42%) during this decade. In comparison, the overall SSLDR increased by a mean of 1.0% per year (95% CI 0.82%–1.23%) from 1.6% in 2010 to 11.6% in 2020 (Figure 2). The increase in SSLDR was dramatically higher than ADR (612% SSLDR vs 128% ADR) over the same decade. Similar trends of ADR and SSLDR were observed when colonoscopies were stratified by patient age and sex (Figures 3 and 4).

Figure 2.:

Adenoma detection rate, sessile serrated lesions detection rate, volume of colonoscopy from 2010–2020. ADR, adenoma detection rate; SSLDR, sessile serrated lesions detection rate.

Figure 3.:

Adenoma detection rate from 2010–2020 based on patient age (a) and sex (b). ADR, adenoma detection rate.

Figure 4.:

Sessile serrated lesions detection rate from 2010–2020 based on patient age (a) and sex (b). SSLDR, sessile serrated lesions detection rate.

Patient characteristics

Overall, we found older patients were associated with a higher ADR (26.2% <50 vs 51.5% >75, P < 0.001) and male patients had a higher ADR than female patients (42.5% male vs 33.1% female, P < 0.001). Patients with a BMI over 30 had ADR of 40.7%, whereas those with a BMI under 25 had ADR of 30.5% (P < 0.001). Lower ADR was observed in Asian patients compared with other races (35.7% vs 38.9% White; 37.8% African American; 36.2% other; and 38.0% unknown, P < 0.001). Current smokers had higher ADR compared with never smokers (46.5% vs 36.9%, P < 0.001). Patients with ≥2 CCI were associated with higher ADR (40.9% vs 37% and 39.4% for those with 1 and 0 CCI, respectively, P < 0.001). High-risk colonoscopies were associated with higher ADR (41.9%) compared with 34.6% for average-risk colonoscopy (P < 0.001) (Table 2).

Table 2. - ADR and SSADR according to patient, procedure, and physician characteristics Characteristics ADR SSADR n/N (%) P value n/N (%) P value Patient-related  Age   <50 yr 2,408/9,208 (26.2) <0.001 847/9,208 (9.2) 0.004   50–75 yr 49,283/130,162 (37.9) 10,697/130,162 (8.2)   >75 yr 3,833/7,448 (51.5) 633/7,448 (8.5)  Sex   Male 31,311/73,645 (42.5) <0.001 5,583/73,645 (7.6) <0.001   Female 24,213/73,173 (33.1) 6,594/73,173 (9.0)  Body mass index   <25 kg/m2 12,965/42,506 (30.5) <0.001 3,168/42,506 (7.5) <0.001   25–30 kg/m2 14,990/40,833 (36.7) 3,076/40,833 (7.5)   >30 kg/m2 9,889/24,276 (40.7) 1,828/24,276 (7.5)   Unknown 17,680/39,153 (45.2) 4,105/39,153 (10.5)  Race   White 31,694/81,473 (38.9) <0.001 7,936/81,473 (9.7) <0.001   Asian 13,944/39,059 (35.7) 2,391/39,059 (6.1)   African American 1,071/2,832 (37.8) 135/2,832 (4.8)   Other 2,168/5,980 (36.3) 349/5,980 (5.8)   Unknown 6,647/17,474 (38.0) 1,366/17,474 (7.8)  Smoking status   Never smoker 37,419/101,419 (36.9) <0.001 8,595/101,419 (8.5) <0.001   Current smoker 2,381/5,126 (46.5) 430/5,126 (8.4)   Former smoker 12,122/28,785 (42.1) 2,388/28,785 (8.3)   Unknown 3,602/11,488 (31.4) 764/11,488 (6.7)  Charlson comorbidity   0 24,014/67,046 (35.8) <0.001 5,946/67,046 (8.9) <0.001   1 10,647/28,801 (37.0) 2,388/28,801 (8.3)   ≥2 20,863/50,971 (40.9) 3,843/50,971 (7.5)  Colonoscopy indication   Average risk 28,342/81,893 (34.6) <0.001 6,282/81,893 (7.7) <0.001   High riska 27,182/64,925 (41.9) 5,895/64,925 (9.1) Procedure-related  Mucosal assist device   No 44,932/124,195 (36.2) <0.001 9,523/124,195 (7.7) <0.001   Yes 10,592/22,623 (46.8) 2,654/22,623 (11.7)  High-definition colonoscope   No 102,38/30,289 (33.8) <0.001 1,384/30,289 (4.6) <0.001   Yes 34,962/81,021 (43.2) 7,972/81,021 (9.8)   Unknown 10,324/35,508 (29.1) 2,821/35,508 (7.9)  Withdrawal time   <6 min 122/903 (13.5) <0.001 14/903 (1.6) <0.001   6–9 min 6,662/25,608 (26.0) 802/25,608 (3.1)   >9 min 38,519/84,644 (45.5) 8,658/84,644 (10.2)   Missing 10,221/35,663 (28.7) 2,703/35,663 (7.7)  Preparation quality   Adequate (excellent, good, or fair) 55,063/145,241 (37.9) <0.001 12,078/145,241 (8.3) 0.004   Inadequate (poor) 461/1,577 (29.2) 99/1,577 (6.3)  Sedation type   Moderate sedation 52,123/139,056 (37.5) <0.001 11,426/139,056 (8.2) <0.001   Other 3,401/7,762 (43.8) 751/7,762 (9.7) Physician-related  Endoscopist sex   Male physician 41,072/110,018 (37.3) <0.001 8,306/110,018 (7.5) <0.001   Female physician 14,452/36,800 (39.3) 3,871/36,800 (10.5)  Years in practice   <11 yr 18,701/44,765 (41.8) <0.001 4,876/44,765 (10.9) <0.001   11–25 yr 29,574/77,299 (38.3) 6,435/77,299 (8.3)   >25 yr 7,249/24,754 (29.3) 866/24,754 (3.5)

ADR, adenoma detection rate; SSADR, sessile serrated adenoma detection rate.

aHigh risk includes those with personal history of colon polyps, family history of colon cancer, and family history of colon polyps.

Younger patients were associated with a higher SSLDR (9.2% <50 vs 8.5% >75, P < 0.001). SSLDRs were higher among female patients compared with male patients (9.0% vs 7.6%, P < 0.001). Patients with unknown BMI had a higher SSLDR (10.5% vs 7.5% known BMI, P < 0.001), but SSLDRs were comparable among the known BMI groups (P = 0.91). Higher SSLDRs were seen in White patients compared with other races (9.7% vs 6.1% Asian; 4.8% African American; 5.8% other; and 7.8% unknown, P < 0.001) and 1 CCI (8.3% vs 7.6%, 7.5% for those with 0 and ≥2 CCI, respectively, P < 0.001). Lower SSLDR was observed among those with unknown smoking status (6.7% vs current smoker 8.4%; former smoker 8.3%; and never smoker 8.5%, P < 0.001). High-risk colonoscopies also had higher SSLDR compared with average-risk colonoscopies (9.1% vs 7.7%, P < 0.001) (Table 2).

Procedure characteristics

ADR and SSLDR varied by procedure characteristics. Longer cecal withdrawal time was positively associated with ADR and SSLDR (45.5% for >9 minutes vs 26.0% for 6–9 minutes, P < 0.001 for ADR; 10.2% for >9 minutes vs 3.1% for 6–9 minutes, P < 0.001 for SSLDR). The use of mucosal assist device and high-definition colonoscope had higher rates of ADR (46.8% vs 36.2%, P < 0.001; 43.2% vs 33.8%, P < 0.001, respectively) and SSLDR (11.7% vs 7.7%, P < 0.001; 9.8% vs 4.6%, P < 0.001, respectively). Similarly, both ADR and SSLDR were higher in those with adequate preparation vs inadequate preparation (37.9% vs 29.2%, P < 0.001 for ADR; 8.3% vs 6.3%, P < 0.001 for SSLDR). In addition, moderate sedation was associated with lower ADR and SSLDR compared with other types of sedation (37.5% vs 43.8%, P < 0.001 for ADR; 8.2% vs 9.7%, P < 0.001 SSLDR) (Table 2).

Physician characteristics

We found ADR and SSLDR did differ based on physician sex and the years in practice. Female physicians have higher rates of ADR and SSLDR compared with their male counterparts (39.3% vs 37.3%, P < 0.001 for ADR; 10.5% vs 7.5%, P < 0.001 for SSLDR). Furthermore, physicians who have been in practice for a shorter time tend to have higher ADR and SSLDR compared with those who have been in practice for longer (41.8% for <11 years in practice vs 29.3% >25 years in practice, P < 0.001 for ADR; 10.9% for <11 years in practice vs 3.5% >25 years in practice, P < 0.001 for SSLDR) (Table 2).

Multivariate analysis

Multivariate regressions had similar findings (Tables 3 and 4). ADRs were higher among older (>75: OR 2.67, 95% CI 2.44–2.92 compared with <50 years old), higher BMI (>30: OR 1.40, 95% CI 1.33–1.47 compared with BMI <25), smokers (current: OR 1.39, 95% CI 1.29–1.50; former: OR 1.09, 95% CI 1.06–1.13, compared with never smoker), ≥2 CCI (OR 1.14, 95% CI 1.10–1.17 compared with CCI of 0), were high-risk colonoscopy (OR 1.07, 95% CI 1.01–1.13), use of mucosal device (OR 1.18, 95% CI 1.01–1.39), or longer cecal withdrawal time (>9 minutes: OR 4.23, 95% CI 2.84–6.30; compared with <6 minutes). ADRs were lower among females (OR 0.68, 95% CI 0.66–0.70) or inadequate preparation (OR 0.72, 95% CI 0.63–0.83) (Table 3).

Table 3. - ORs of ADR according to patient, procedure, and physician characteristics by all years, 2010 and 2020 Characteristics All yearsa 2010b 2020c P-interactiond OR (95% CI) OR (95% CI) OR (95% CI) Patient-related  Age   <50 yr Reference Reference Reference 0.28   50–75 yr 1.75 (1.62–1.88) 1.17 (0.73–1.88) 1.63 (1.40–1.90)   >75 yr 2.67 (2.44–2.92) 2.11 (1.16–3.82) 2.25 (1.77–2.86)  Sex   Male Reference Reference Reference 0.44   Female 0.68 (0.66–0.70) 0.62 (0.49–0.78) 0.69 (0.63–0.75)  BMI   <25 kg/m2 Reference Reference Reference 0.33   25–30 kg/m2 1.18 (1.15–1.22) 1.03 (0.86–1.22) 1.14 (1.03–1.26)   >30 kg/m2 1.40 (1.33–1.47) 1.16 (0.83–1.63) 1.48 (1.31–1.66)   Unknown 1.67 (1.57–1.78) 2.14 (1.67–2.75) 1.10 (0.99–1.23)  Race   White Reference Reference Reference 0.08   Asian 0.98 (0.90–1.05) 0.83 (0.62–1.09) 0.96 (0.86–1.08)   African American 0.92 (0.85–1.00) 1.25 (0.78–2.01) 0.91 (0.74–1.12)   Other 0.95 (0.89–1.02) 0.82 (0.55–1.22) 1.13 (0.91–1.39)   Unknown 0.98 (0.92–1.04) 0.95 (0.74–1.22) 0.91 (0.82–1.01)  Smoking status   Never smoker Reference Reference Reference 0.08   Current smoker 1.39 (1.29–1.50) 1.32 (0.75–2.29) 1.33 (1.07–1.65)   Former smoker 1.09 (1.06–1.13) 1.27 (0.99–1.63) 0.98 (0.89–1.08)   Unknown 0.94 (0.88–0.99) 0.67 (0.49–0.92) 0.99 (0.86–1.14)  Charlson comorbidity   0 Reference Reference Reference 0.08   1 1.02 (0.99–1.06) 1.24 (1.07–1.43) 1.03 (0.95–1.11)   ≥2 1.14 (1.10–1.17) 1.29 (1.04–1.59) 1.15 (1.05–1.26)  Colonoscopy indication   Average risk Reference Reference Reference 0.73   High risk 1.07 (1.01–1.13) 1.33 (0.91–1.95) 1.24 (1.11–1.39) Procedure-related  Mucosal assist device   No Reference Reference Reference NAe   Yes 1.18 (1.01–1.39) NAe 0.92 (0.76–1.11)  High-definition colonoscope   No Reference Reference Reference   Yes 1.08 (0.97–1.19) 0.76 (0.52–1.11) 0.97 (0.82–1.15) 0.26   Unknown 1.04 (0.77–1.40) 0.86 (0.24–3.07) 1.08 (0.84–1.38)  Withdrawal time   <6 min Reference Reference Reference 0.10   6–9 min 1.85 (1.29–2.65) 2.54 (1.31–4.92) 1.27 (0.75–2.16)   >9 min 4.23 (2.84–6.30) 6.42 (2.73–15.12) 2.46 (1.59–3.82)   Missing